ML Engineer · Building AI Agents that think, plan, and execute.
USC MS CS '25 · Ex-Cadence · Ex-Persistent
name: Deep Vivek Sheth
location: San Francisco, CA
education: MS Computer Science @ USC (2025)
current_role: Building AI systems that reason, plan, and act
interests:
- AI Agents & Multi-Step Reasoning
- GPU-Accelerated Computing (CUDA)
- Distributed Systems & Database Internals
- Full-Stack Product Development
fun_fact: I optimize everything—code, travel routes, and coffee brewing time ☕Languages
AI/ML
Infrastructure
- Multi-Agent Systems — Coordinating multiple AI agents for complex tasks
- vLLM & Inference Optimization — Making LLMs run fast and cheap
- Advanced RAG Patterns — HyDE, RAPTOR, and Agentic RAG
Odyssey — AI-Powered Travel Planner with Zero Backtracking
GPT-4 Agents · Google Maps API · RAG · Next.js
"Never drive back and forth again." A multi-step AI agent that plans geographically optimized day trips using real-time distance calculations and review synthesis.
🗺️ Odyssey — AI Travel Agent
Full-stack AI planner that generates optimized itineraries using GPT-4 agents, Google Places API, and geospatial analytics. Built with FastAPI + Next.js.
Highlights:
- Multi-step agent workflows with ReAct pattern
- Real-time Distance Matrix integration for route optimization
- Review synthesis for "vibe-aware" recommendations
| Project | Description | Tech | Status |
|---|---|---|---|
| 🤖 Automated_SRE | Agentic AI SRE with RAG + Gemini | Python, FastAPI, ChromaDB |  |
| 📜 LogParser_LLM | 33,000x cost reduction log parsing | Python, Gemini, Prefix Tree | |
| 🏥 unity-companion | AI Patient Results Assistant | Next.js, FastAPI, RAG | |
| 🗺️ Odyssey | AI-powered travel planner | TypeScript, FastAPI, GPT-4 | |
| ⚡ Parallel PageRank | 300x GPU speedup | C++, CUDA, MPI | |
| 🗄️ ADS Buffer Manager | PostgreSQL buffer strategies | C, Spinlocks | |
| 🔍 RAG-Search | RAG with FastAPI & Next.js | Python, LangChain | |
| 🎓 LLM Teaching Assistant | Agentic course assistant | Python, OpenAI | |
| 🤖 ChatGPT from Scratch | Built GPT from scratch | PyTorch, Jupyter | |
| 📈 Stock Prediction | News sentiment analysis | Python, LSTM | |
| 💻 LeetCode | 300+ solved problems | DSA |
- 🎓 MS Computer Science — University of Southern California (2025)
- 💼 MLE Intern @ Cadence — Built GPU-accelerated pipelines & RAG systems
- 📄 Published Research — ML-based intrusion detection (IEEE)
- 🧠 10M+ Logs Fine-tuned — Production LLM optimization
- 📧 deepsheth3@gmail.com
- 🌐 Portfolio: Coming Soon
"Ship fast. Measure everything. Iterate."